AIBullisharXiv – CS AI · 5h ago6/10
🧠
PromptEcho: Annotation-Free Reward from Vision-Language Models for Text-to-Image Reinforcement Learning
Researchers introduce PromptEcho, a novel reward construction method for improving text-to-image model training that requires no human annotation or model fine-tuning. By leveraging frozen vision-language models to compute token-level alignment scores, the approach achieves significant performance gains on multiple benchmarks while remaining computationally efficient.